Intro. [Recording date: June 20, 2023.]
Russ Roberts: Today is June 20th, 2023. My guest is author and economist, Daron Acemoglu of MIT [Massachusetts Institute of Technology]. This is Daron’s sixth appearance on EconTalk–last here in September of 2019 talking about shared prosperity and good jobs.
Our topic today is technology. We’ll be referring to his new book with Simon Johnson, Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity along with some other work that Daron has done on Twitter. Daron, welcome back to EconTalk.
Daron Acemoglu: Thank you, Russ. It’s a true pleasure to be here with you.
Russ Roberts: Now, unlike many economists, you and Simon Johnson have come out more worried about technology, I think, than in the usual mainstream. You are particularly concerned about living standards, generally, wages; and you raise the question as to whether technology will necessarily improve living standards widely. Your book takes a long, broad look at very historical episodes. But make the case: Why shouldn’t I trust technology to lead to making people better off? Which is a very–I mean, it’s almost a religious view for me, from my training.
Daron Acemoglu: Thank you, Russ, and you summarized it brilliantly, especially with the word ‘necessarily.’ And that’s really the main reason why we wrote the book.
And, Simon and I are convinced that–as economists and as concerned public policy makers, analysts–we have to hold two potentially conflicting views in our mind. One, that today we are extremely fortunate, from a historical sweep, because of the technological breakthroughs that started sometime around the middle of the 18th century. The application of knowledge–later scientific knowledge to industrial processes–that have brought us much better health, much greater comfort, much greater level of prosperity. So, we are huge beneficiaries of technological progress.
But, at the same time, we also have to hold the conflicting thought with that: that technology can be used, and has been used for bad; and there is nothing automatic about technological progress bringing widespread prosperity.
And that automatic–or ‘necessarily,’ as you put it–is the key thing. Because in thinking that somehow technological progress is going to be such a powerful force for lifting all boats automatically, we abrogate our responsibility to shepherd it in the right direction and to make institutional adjustments so that society at-large benefits.
And, the reason why we do all this historical work in the book is to argue that we have benefited from technology precisely when there are certain preconditions that have been satisfied.
Russ Roberts: So, my take is–I would argue–your position is half right, in my view, and I want to give you a chance to convince me that you’re completely right. It’s clear to me that our standard of living is wildly better than the standard of living of people 250, 200, 300 years ago, as you suggested.
I think it’s difficult to make the case that the enormously widespread availability of good living standards was shepherded in any way.
So, competition is a very powerful force. It’s true that the innovator–the person who develops the technology–can sometimes capture a disproportionate large share of it initially. But, over time, competition from entrance lowers the price, creates access, and raises living standards. It also saves resources to allow new things to be created.
Now, where I would agree with you is it certainly doesn’t raise all boats at every point in time. Right? Certainly technology that threatens the–that competes with people already doing something that is well paid and now is suddenly not going to be well paid, it’s going to hurt those people in the short run. And their short run may be half a lifetime, which is very painful.
But, I don’t understand the shepherding argument. I don’t see much in the history of technology that suggests it’s been shepherded much. Change my mind.
Daron Acemoglu: Well, I don’t know whether I can change your mind, but what I’m going to do is I’m going to first claim three critical ingredients for technology to have the type of effect that you explicated. And then, I’ll give you some historical examples. And of course, feel free to interject. I don’t want to go on for too long.
But One is: competition. Absolutely, you are 100% right. But, I think you are not right in thinking that competition is automatic, either in our age or throughout history.
Second: coercion in labor markets. So, that’s the tip of the iceberg in some sense. That means that when employers have too much power over their employees, there is no guarantee that an increase in marginal productivity of labor–okay; sorry, I’m using a jargon, but I think it’s okay for an EconTalk–an increase in the contribution of labor to productivity will translate into higher wages.
And, Third: automation. That, by itself automation may increase capital’s productivity–may increase profits–but it may not increase the marginal productivity of labor; and hence break one of the chains that goes from higher output, higher productivity to higher wages.
We see all three of them in history. So, we spent quite a bit of time on medieval technologies to argue that the so-called Dark Ages weren’t dark in terms of technology. But they were dark in terms of the conditions of the peasants. And, part of the reason was this lack of competition and coercion both of the first two factors I’ve mentioned.
Windmills, for example, increased productivity more than 10-folds in some tasks, but because they were monopolized by lords and abbots who actually forcefully did not even allow any competition and coerced people to use their mills, few, if any of the benefits spread beyond that small, narrow elite.
Let me give you another example of coercion. A transformative technology for the United States’ history is the cotton gin. It turned the United States from–it’s essentially a backwater for agricultural productivity, especially in industrially relevant agriculture–into the largest exporter of cotton. It revolutionized the South’s economy.
But, the workers–black, enslaved people–actually fared much worse under the cotton gin because they were moved into much, much harsher plantations in the Deep South, as cotton plantations spread.
And then, finally, automation. When we see automation being the only or the main focus of technological changes, we see that many workers experienced lower wages or lose their jobs. That’s what happened in the first phase of the Industrial Revolution. It went longer than half a lifetime. It may have gone on for about 90 years or so–perhaps two lifetimes for people at the time, especially since life expectancy at birth declined to around 30 years as people moved into the cities.
So, all three of these are important. And when I meant ‘shepherding,’ I meant ensuring competition, ensuring labor has voice and power, and also encouraging broader applications of technology than just automation.
Russ Roberts: So, let’s concede–I think we agree on competition, not surprising–where we might struggle. I’ll give you a chance. You can talk about it if you want. But, I think the challenge is how do you bring about competition?
One of the biggest challenges, of course, is that it’s not so much that there is a competition. It’s that there’s barriers to competition. Some of those may be natural; many of them are often government-imposed. And we probably agree that we should probably get rid of those.
The challenge, of course, is figuring out which ones those are. Because, in our current world, as we’ll get it to later, technological–the biggest corporate power in the United States, the most successful companies–in theory, there are no barriers to entry, but they seem to have a pretty good deal for themselves. And, we’ll get to that when we talk about ChatGPT [Chat Generative Pre-trained Transformer], I think.
Russ Roberts: Let’s move on to the other two–the issue of labor coercion. And, what was the third one again?
Daron Acemoglu: Automation versus other things that increase marginal productivity.
Russ Roberts: Okay.
So, on the case of labor, this is a growing concern, I think among many economists, that what is often called bargaining power is inadequate to allow workers to benefit from the pie getting larger, say.
I don’t understand that. Let me ask a more general question. You talked about the Industrial Revolution. Obviously, there were many parts of the Industrial Revolution that were unpleasant. Very similar to today’s world in, say, China, where a lot of workers are–their activity is not well measured. They’re in rural areas; they’re not part of the data. They come to the cities. City life is challenging. It’s culturally challenging. They struggle; and many of them may be worse off in the short run and it may be a fairly long time.
There’s a big debate. And I just want to give some credence to both sides. I think I know where you stand on it. But, a huge debate in economic history as to how the Industrial Revolution treated workers. So, the rough period we’re talking about here is something like 1750 to 1880, particularly in England.
And, it’s England it gets looked at, because it’s the beginning of it. It’s where we have the most data. And, I would argue that time period is pretty imperfectly understood.
There are very, very loud views on both sides of this issue. There are people who say, ‘For 90 years, 100 years, no progress.’ You say it in your book: ‘No progress for the average worker.’ Other people disagree. They say, ‘Actually there was quite a bit of progress.’ Now, there may be pockets that didn’t gain. But, how confident are you of your view of that? Because it’s somewhat important to your argument, it seems.
Daron Acemoglu: Absolutely.
Well, actually, Russ, you’ve made two points. One is about coercion more generally. And, I want to come back to that because I think it’s very, very important.
But, if we can put a pin on that–and let’s not forget it–let me answer your question about the Industrial Revolution. No, I am certainly not confident about real wage data between 1750 and, say, around 1840. And, there are voices on both sides. You are absolutely right.
But the general statement that the working people did not much benefit from the industrial progress, I think I’m fairly confident. Because, there you have to bring in what was happening to their living conditions? What was happening to their work autonomy? What was happening to their health? What was happening to their children?
I think when you look at the data there, to me there is very little doubt. In 1840 [1834–Econlib Ed.], there was a Parliamentary Report about the condition of England that reported children as young as five or six working 18 hours in deep coal mines and pushing carts with their heads. 1840. 1840. This is 90 years after the 1750, which is the rough beginning date of the Industrial Revolution.
1720 is the beginning of first big textile mills. So, maybe under 120 years after that.
And, the reason why we have these data is because the British had parliamentary commissions; and there was starting–why so late? That’s a question–was starting some concerns about the conditions of England, because of petitions, because of people talking and newspapers.
So, this is the condition that we’re looking at. 1850s: Still, the living conditions in British cities–where more and more people were living–were horrible. Life expectancy may have fallen below 30 years at birth. Infectious diseases were rampant. There was no sewage system. And, even though we don’t know about real wages, working hours expanded quite significantly under some excellent economic history works that really document this.
So, yes, we can debate about real wages, and it’s a good debate. But, the conditions that we’re talking about were really very, very different from what started to transpiring from around the 1860s, 1870s.
And one important factor there is, you know, worker voice. Trade union activity was very harshly suppressed in Britain. Any kind of democratic movement was super-strongly discouraged.
1840s, when I was just mentioning a second ago, it was also the beginning of the Chartist Movement. Chartists collected 3 million signatures. I mean, let that number sink. Three million signatures for asking for minimal changes in the political system so that people could vote and could participate in the political process. They were very careful not to make even demands about trade union organization so that they could not be labeled Socialists.
And you know what? The British upper-middle classes or middle classes–or whatever you want to call the system–the people who were politically powerful, they completely ignored them and threw all of the Chartist leaders in jail.
So, that’s what I mean that there was nothing automatic about that process.
And when Trade Unions start, you know, organizing after the Master and Servant Acts–which completely disempowered workers against their bosses, and Trade Unions were legalized–that’s when you see conditions in factories improved quite a bit.
Russ Roberts: So, let me ask a more general question, then, related to that. I certainly don’t deny that life was very hard for many people in this time period. I think the idea of using standard-of-living or real wages is an attempt to get some objective measure that is somewhat across a wide part of the population.
Let’s think about the United States, and immigrants–come to the United States in the second half of the 19th century, in very large numbers. And, they come to very horrible conditions. They come to tenements in New York City, where as you say, disease is rampant. There’s huge overcrowding. There’s very many bad things in terms of quality of life, especially compared to where they had come from.
So, you might have to ask yourself: Why did they come? Were they misinformed about the life they were living?
Similarly, the people who flocked into London in and Manchester and the cities that were industrializing in the last half of the–in the whole 19th century in England–last half of the 18th century: Why would they come? Why didn’t they just go–when it was horrible and life was tough? And, working in the cold, [inaudible 00:17:20? coal mill?] with your head. At the age of six, your kids having a horrible day. It is–it’s horrifying.
Why didn’t they go back somewhere else? Why didn’t they go back to the countryside? Why did immigrants come to the United States if it was–it was horrible–on the surface?
My contention would be maybe it wasn’t as bad as it looks to us today, but more importantly, maybe they thought there was a future for their children or grandchildren? It turned out to be correct in the case of the United States. Maybe not for England. Maybe it took their great-grandchildren, or great-great grandchildren; and it could have come sooner with the right regulation. So, I think that’s the crucial question. But, first just react to that, that overall question.
Daron Acemoglu: Yeah, you nailed it. You nailed it. I think–exactly. It could have come sooner, and it did come. But again, it’s not the automatic process.
And, I’ll come back to the United States–because I want to come back to the United States in the context of automation issue because I think it’s a great example.
But just as a preliminary answer: Yes, they migrated to the United States because living conditions were better in the United States. Wages were higher in the United States. Wages are key.
And, why wages were higher? Well, that’s about labor shortages and technology, both of them; and we’ll come back to that.
But, the reason why they went to London is mostly push factors. So, labor was pushed out of the countryside in the United Kingdom for two reasons. One of them is semi-good: Agricultural productivity was increasing in a way that reduced labor requirements, so people did not have as much work.
But second, there was also this whole Enclosure Movement, which rationalized parts of agriculture, but in a very harsh way, cutting out many of the means that people had for surviving. Because, in the open fields, they could collect wood, they could hunt, they could do other things in order to support their existence there. And again, that sort of illustrates some of the tensions that we’re talking about.
One of the, sort of, most eloquent observers of the Enclosure Movement was Arthur Young, who started as, you know, almost like a Malthusian before Malthus in terms of his views, and was very–very much–an economist by instinct–thinking about economic efficiency and how important it was to introduce rational property rights, etc.
But, after a while, he changed his mind and he said, ‘Look, through this Enclosure Movement, we’re really treating these people so badly and we don’t need to do it in this harsh way–in this rapid way–that throws them out. And we could at the same time allow them some means of existence.’
You know, again, this is sort of the tension. We are all recognizing that the economy had to be transformed.
And technology was providing the means.
But, how are we going to use these means? And through what institutional means?
And, that’s exactly why you nailed it. With the right kind of regulation, it didn’t need to be three generations. It could have been perhaps one generation in the United Kingdom.
Russ Roberts: But, what’s the evidence that the quicker pace in the United States was due to regulation as opposed to just other factors? I mean, there’s nothing–it’s not that–I don’t think there’s any evidence of that, that the United States–
Daron Acemoglu: I’m not claiming that–in the United States, it was regulation. I think the evidence for, quote/unquote–“regulation”–is not even the right word. I just picked that one because you used it.
The institutional framework, is: Is it Trade Unions? Once Trade Unions were in the mix, then you see conditions in factories improve, and worker voice, more investment in education, more investment in public services. So, that’s the sort of the institutional balance.
And, more generally democracy.
You know, the United Kingdom at one level–and this is what for example, in my work with Jim Robinson, we’ve celebrated this–is a huge success of democracy, a peaceful process of democratization–but, on the other hand, it was a slow one. It took, you know, again, 1860s, 1870s, until really truly there was some sort of democracy which provided voice to the working people.
Russ Roberts: Yeah, I think that’s a bit of an illusion. I don’t know if we want to spend a lot of time on that in the sense that I’m not sure that the political power of those groups made a large difference in, say, standard living and social services or other things. There were very low levels during that period.
But, I want to make your point a little more stronger because I think you correct me–wisely. You know, I said, ‘Why did they come to the United States?’ or ‘Why did they go from foreign countries–why did they go to the industrial centers, manufacturing centers of England?–if it wasn’t a better life?’
And, the answer is, ‘Well, it could have been.’ That’s one of the reasons.
But, the other reason is it could have been so bad where they were. And, your point about immigration is certainly true. Jews fleeing Czarist Russia, pogroms, antisemitism; Irish fleeing of potato famine. Many, many immigrants in the United States came because–not because it was glorious and the streets were in fact paved with gold, but because it was the alternative [inaudible 00:22:43]–
Daron Acemoglu: They even came from England. That’s exactly your point: they came from England.
Russ Roberts: But, let’s talk about the trade unions. Because, I was–in my day as an economist, when I was a real economist in grad school in the 1970s, late 1970s, the consensus was that trade unions had no impact on wages. They were ineffective. They verified–it’s not verified, I forget the word I want–they confirmed changes in wages that otherwise would’ve happened. The only way they can actually affect wages is by contracting employment.
So, while they can raise wages in certain particular industries, they do that by reducing employment in those industries. And so, the net effect on workers is zero at best.
So–and then I look at the last half of the 20th century: Industrial Standard of living is rising steadily in the United States as unionization steadily falling.
So, it seems like a lot evidence–from my very biased perspective–that unions are not important for sharing the pie. They rearrange it a little bit, but otherwise, not really what makes the difference. It’s actually just the steady growth in technology–application of capital–and the things that you’re suggesting need to be stewarded.
Daron Acemoglu: So, let me say a couple of things, Russ. And, again, I still want to come to technology in the United States because that’s the critical part for the United States and it’s also a critical part of our argument.
But, first of all, long-run effects of trade unions are very difficult to know. So, if you have a very strong opinion about long-run effects–like, in 30 years time, how will they change wages, etc.?–that’s very difficult to nail.
But, short- to medium-run, I think there are both good theoretical reasons and empirical evidence that when they are effective–and that effectiveness requires both trade unions to behave the right way and the employers to react to them the right way–they can have quite significant effects.
In terms of the theory, just very quickly: You know, the question is whether you start from a labor market where everything is fully competitive except the trade union, or whether you start with a frictional labor market. Like, for example, if you have a frictional labor market where there are, for instance, search frictions–which mean that every employment relation has a quasi-rent in it. Meaning that: I would love to be here and you want me to be here, Russ, and it’s not like I have an outside option that will make me exactly indifferent. And, you have an outside option at this point that will make you exactly indifferent.
I think in the medium run, that’s right. If you planned for next week, you could find as good a guess as me. But, right now we are locked into this relationship. That means that in our relationship there’s a quasi-rent and we have to distribute it. How are we going to distribute it? That’s the bargaining problem. And, if that bargaining for whatever reason–for instance, if I’m an enslaved worker or the rest of the economy is treating me really badly–you have all the bargaining power in there, that’s actually not going to lead to a fair division. It’s going to lead to inequality.
Actually, it won’t necessarily lead to much efficiency, either.
So, that’s the question, which is: How are these employment-generated quasi-rents distributed?
And that becomes a particularly big problem when firms get bigger like the modern factories because they have a lot of power over their workers.
And the way to think about why trade unions did have a meaningful effect, and without creating huge inefficiency–again, that may be very different than from the police unions and the teacher unions–but, in the modern factory system, it’s because they level the playing field to some degree.
And where do you see the evidence for that?
In the United States it’s more complicated, but if you look at Scandinavian countries, there you are actually seeing how unions, when they are acting in a way that’s in concord with employers, but actually, you know, asking for better working conditions or better wages, also compressing the wage structure, they are having massive effects that you can see in the data.
Russ Roberts: Well, I don’t know those data. That’s interesting. And we’ll link to anything you share, to let people explore it. I don’t understand the argument that a large firm, say, has more power over me. You and I have spent our lives in the educational system. They are nonprofits–our employers–typically. And, I always like to make the argument: That doesn’t mean they’re nice. We don’t make the wages we make because our employers are nice, because they’re nonprofit. We could probably–at least, I know you could; maybe I could–we could make more money elsewhere in profit-oriented places, perhaps Wall Street or elsewhere.
So, we have plenty of market power, you and I. Just one person. And, why? It’s simple: We have alternatives. And, those alternatives–there’s no bargaining. The wage that economists make is–I’m not going to try to justify it to some perfect competition system, but it is determined by how much opportunity we have elsewhere. It certainly has–my willingness to complain, has something to do with it. There’s a little bit of bargaining. But, in general, the wages are set. They’re posted. They’re not bargained for. They’re the result of a quasi-market process.
And, the large employer–forget our own employers–large employers in the United States. Let’s take Walmart, which is a perfect example: I think the largest private sector employer in the United States. They have to set wages to attract people away from other retailers and other places that use relatively low-skilled labor. At Walmart, the skills are showing up on time and there are things that you can acquire or have access to. Do you believe that they are able to exploit workers or charge less because they’re bigger than they were when they first started?
Daron Acemoglu: Well, there are actually two issues here, Russ. I think there are two issues and they’re both very important. And, again, some of these are not as central for our arguments, but I think it’s useful to talk through them. One is monopsony power. And, I am not one of the new converts [?converse?] to monopsony power, meaning that I think there are instances of monopsony power, but–
Russ Roberts: Explain what monopsony is for listeners who are not–
Daron Acemoglu: So, monopsony–here is the example of monopsony that I’ll give you, which I think is an answer to one of your questions; and then I’ll give you the caveats about it. If you have a company town, that is a monopsony. There is one employer who organizes the town and controls not just, say, industrial production there, but also the service jobs. If you disagree with that employer and he says, ‘I’m going to fire you,’ there’s no other job for you. The only thing you can do is you have to move out of that city. And, that moving is very costly. So, that is monopsony power, meaning that he is the single employer and he can, under some conditions, exploit that by pushing down wages and moving down your labor supply curve. And, that’s the classic monopsony model that John Robinson and others have worked out.
Now the question–so, there’s no doubt in my mind that is relevant for certain historical episodes. The question is, A. whether that is relevant for the United States labor markets today. And, B. whether that is the most important labor market imperfection in the United States. On B., I’m convinced the answer is no, meaning that–and that’s why I started from a different model of how the labor market works. And, that’s what I meant by the new converts [?converse?] to monopsony. A lot of people have now turned to thinking of monopsony as the organizing framework for labor market imperfections; and I’ll come back to why.
For my thinking more generally, a little bit for the book with Simon, that’s not the right sort of thing. I think that’s important for me to explicate that. I don’t think I have explained it anywhere else. On A., I think it’s an empirical question. So, you again sort of nailed it with your Walmart question, which is: Does Walmart have monopsony power?
And the answer is: It depends. If Walmart is the only employer for low-skill workers, yes, it would have monopsony power. But, if there are other jobs that are not as bad and a variety of low-skill workers could perform–cleaning, security work, customer-facing industries in the kitchen of McDonald’s, etc.–it’s going to depend. So, it’s going to depend on the elasticities with which they could switch comparative advantage in one task versus another. So, it’s going to be an empirical question.
So, my reading of the evidence is that in the labor market today, there is some–in the United States–there is some amount of monopsony power. But, I still come back to the B., which is I don’t think that’s the main reason why labor markets are imperfect.
And, in fact, that’s why I sometimes think we shouldn’t overdo monopsony power. Why do I think? Because I think the relevant labor market imperfection, I think it’s very, very important in my thinking is the other one that I started with. There’s a quasi-rent between you and me, meaning that you are–it’s going to be very costly for you to fire me and find another guest, and it’s going to be costly for me to say, ‘I’m going to cut this interview in half. I’m not continuing.’
So then the question is: How do we in general split those gains? And, whether that’s a cooperative process or versus a non-cooperative process.
And, the type of worker voice to me that’s most conducive is the one that makes sure that I can protect my interest, but this relationship doesn’t become very conflictual. And when that happens, and when employers also have the right attitude, then you have things like gift exchange, meaning employers do favors to their employees, they have a view of employees as part of the stakeholders, part of the family, and employees then make adjustments so that they can be more secure. They make the training investments and other investments. But it is all in this context of splitting these quasi-rents.
I think that’s a very different view than the monopsony model. It has some implications that are similar. It has some very different implications. For example, ensuring cooperation between employer and employee is much more conducive in the context of splitting the quasi-rents and having the right sort of balance is important.
So, why do I think large companies are a problem? Because if you have large companies that are very well organized and completely disorganized workers, it’s going to be in the splitting of this quasi-rent that there’s going to be an imbalance. When you are against a bureaucratic organization as an individual, you may become powerless. So, that’s why–and, this is I think what Galbrath was after in his countervailing powers–large corporations, you need some sort of coordination among the workers as well. Otherwise, divide and rule: You are just one individual. You have to follow the harsh conditions that I’m setting. I think that could be one path that’s dangerous.
Russ Roberts: I don’t agree with that, other than in the company town example. And even that, I think people leave town. In a modern industrial economy, I don’t care how big the company is. Do you think the largest employers of engineers in America–the tech sector–as it’s gotten larger, is somehow more able to–they’re desperate for workers. They give them all kinds of amenities. And there’s none of this quasi-rent negotiation. Again, it’s just a posted price. It seems–
Daron Acemoglu: Oh, come on. No, no, no, no, Russ. There, I’m going to disagree with you this time.
Russ Roberts: Please.
Daron Acemoglu: The largest corporations, it’s all quasi-rent negotiations. You know, why is it that in the heyday of the tech sector, companies like Google, Facebook, were doing all of these, you know, amazing things to make employees happy–to give them free snacks, coffees, ‘Come here and we’ll entertain you’? That was all part of splitting the quasi-rents in a way that made the employees happy so that they could build this cooperative, sort of, feeling.
Now, in terms of the low-skill labor markets, the McDonald–
Russ Roberts: Wait a minute. But wait a minute, wait a minute. So, those large firms, they’re still doing it even though they’re much larger. Are you going to argue that wage–
Daron Acemoglu: No, no. But there, they’re dealing with workers that have very specific expertise. There are very useful workers that they didn’t want to lose. But, the reason why they didn’t want to lose them–so, it’s because they’re making quasi-rent.
So, you can’t think of that as a competitive labor market. In a competitive labor market, I would never throw money at you like that because I’m just paying for your marginal product anyway. So, the fact that I am, like, throwing money left, right, and center is already sort of a sign that there are some quasi-rents being split there.
Now, when it comes to the low-skill labor markets just want–let me just point this–
Russ Roberts: Go ahead–
Daron Acemoglu: Then you are absolutely right. It’s an empirical question. Really, these low-skill workers, were they so beholden to McDonald’s and Walmart, or did they have other options?
And, my view is somewhere between you and the new monopsony converts. In some cases, yes. In some cases no. But, it’s an empirical question. There’s some very good empirical papers there that do estimate monopsony power, but the monopsony power is not, like, crazy, like 10% wage markdowns, or so on.
Russ Roberts: So, we’ll agree to disagree on this, to some extent.
I think where we’d agree is that all of these transactions between employers and employees are embedded in a larger labor market. And I’m just pushing the argument that I think that larger labor market is much more protecting of workers than your view. We could disagree on that empirically. We maybe disagree on it theoretically.
But I think of, say, people who are house cleaners who make two to three times the minimum wage. They’re vulnerable because of immigration status often, and yet they’re treated quite well–financially. I’m not saying it’s a pleasant job, but they make much more than the law requires.
And, it’s clear to me why. Because if you don’t pay them that market wage–which is the going wage in the area where you live–they don’t show up because there are many alternatives. Even though those alternatives are just the same in other people’s houses, there are many, many employers. That’s consistent with both of our views of competition. But, the fact that I have, in theory, a huge edge over them because their English is mediocre, there are plenty of jobs they can’t get–maybe because of their legal status. And I’m still paying them many, many times the minimum wage, not because I’m a nice person. Because the market requires it.
Daron Acemoglu: Yes. Um, you know, the United States has had some labor shortages. When labor shortages are acute, that’s exactly what happens in the house cleaning sector. I think it seems like when labor shortages aren’t that acute, then things don’t work out so well.
Russ Roberts: Let’s–I want to move on to–
Daron Acemoglu: And, the labor shortages are actually relevant for the United States case. So, that you had asked earlier on, so why were people coming to the United States?
Russ Roberts: Yeah.
Daron Acemoglu: So, but, if could, if I may–
Russ Roberts: Yeah–
Daron Acemoglu: I want to sort of talk about the third element of that thing: the sort of: why is it, when is it that technological change will lead to prosperity? Because it’s a central part of my work and central part of the thesis of the book, and I think it sort of puts the labor shortage into much greater perspective.
So, again, if I’m going to introduce a little bit of jargon–and this part for those who don’t want the jargon, cover your ears–but, the economist’s view of technology is partially colored by the most convenient models we use–such as, for example, Cobb-Douglas production function. And, the Cobb-Douglas production function is special, and you don’t need to know what it is. But, what it implies is that marginal product and average product are proportion. Which means that whenever something increases the average productivity of a firm or an industry–meaning that how much output is produced per worker–that also means that exactly the contribution of the worker is marginal productivity increases by the same amount.
So, in that world, technology is going to be wonderful for workers. Technology improves the average productivity, which means how much output I produce as a firm with the same number of employees by twofold. The equilibrium is going to lead to twofold increase in wage.
The world is not like that, especially when there are a lot of automation technologies. And if you want to think about it that way–we discuss this as an illustration in the book as well–sort of this claim about the factory of the future, which is ascribed to many people; so, let me not give it to any one person–is that the factory of the future will have two employees, a man and a dog. The dog is there to make sure that the man doesn’t touch the equipment and the man is there to feed the dog. So, obviously, it’s an exaggeration, but if that’s the factor–
Russ Roberts: It’s close–
Daron Acemoglu: It’s close–
Russ Roberts: There are many, many productive processes in the United States where it’s such a trivial number of workers relative to output.
Daron Acemoglu: But, the reason why this is such a good example is that it clearly highlights why that man–or many people who may be working in these companies–don’t really contribute to average productivity in that huge way. You can eliminate this person and the dog, and the factory would still work fine. When that’s the case, the labor market–the competitive process–shouldn’t pay this person a high wage. That’s automation.
And, the key that my work has done in pushing is two ideas. And, the book with Simon develops these further and provides a lot of historical background. One is that there shouldn’t be any presumption that just automation is necessarily going to lead to much higher wages.
But second, that we don’t need to do just automation. So, this is where the shepherding is important. Technology, human knowledge, scientific knowledge is a very broad platform. We can use them in many different ways. We can create different future factories where humans have much greater contributions in terms of design, maintenance, flexibility, social skills, creativity. And, if we do that, then we’re going to increase wages.
Now, if we come back to the United States, why was the United States so attractive to low-skill workers? This is actually something that the economic historian, H.J. Habakkuk, first hypothesized and provided some qualitative evidence for. Because, skilled labor was in very short supply in the United States, the whole creative energy of American innovators and entrepreneurs was devoted to increase the productivity of unskilled labor.
So, unskilled labor was centrally integrated into the production process. So, in the United Kingdom and the United States–and the United Kingdom again sent a lot of Parliamentary commissions and they said exactly the same thing. So, we have great qualitative, detailed description of this. So, in the United Kingdom, you had these artisans who were trying to work with a chisel with very good hand-eye coordination in order to make manufacturing processes work. Whereas the Americans said, ‘We’re not going to find these skilled artisans, and even if we do, they’re going to be very expensive. So, let’s introduce machinery in a way that can be used with unskilled labor.’
That was the beginning of the interchangeable parts system, the American system of manufacturing. And, once that got off the ground, unskilled labor was very central to the manufacturing, and its productivity was very high, and its wages were very high.
Russ Roberts: The part that’s missing from your story for me is that–let’s take the man and the dog. And I’ve written about this a little bit. The many, many processes in America in modern industrial–it’s not just in America: modern industrial societies–they’ve stripped away immense amounts of labor through the substitution of machines. And you’re right, it doesn’t make the workers who are left productive, because they have very little–the process is productive. The total output divided by the number of workers is very high. But, because the skills of the workers are relatively low–which is the whole idea of the process–they don’t make a high wage. They don’t capture any of the productivity gains that the machinery brings.
So, I’ve written about this in pencil manufacturing, egg farming. The modern egg farm is an enormous number of chickens and a handful of people–a couple. And, their main job is to take dead chickens out. And, like you say, to make sure–the whole thing is very–to make sure that the machines don’t break. But, as long as they’re working fine, they have very little role, and I’m pretty confident they don’t make a lot of money. The money mainly goes to the people who bought the machines, installed them, the people who made the machines and created them.
However, the net result is an enormous drop in the price of eggs. And, that means that the workers who work elsewhere have a much higher standard of living, including the ones that work in that plant. And that’s happening all over the economy.
And so, what’s happening is–here’s the irony–as the innovation is stripping out labor from many different manufacturing processes, that’s creating opportunities for new employers to find things that those low-skill workers could do, and they have. Historically, there isn’t mass unemployment in the face of innovation. And the whole 20th century in the United States, is that story, to me.
Maybe I’m missing something? It’s the application of technology and automation to things that used to be done mainly by hand so that the hands aren’t necessary; and those hands go elsewhere and they work in other parts of the economy, and they’re able to enjoy a decent or even much higher standard of living because many of the things they buy are now less expensive because of that technology. And, they don’t want to share those gains, the owners of the egg farming factory, but they’re forced to because there’s too many people making eggs.
Daron Acemoglu: Russ, you gave an excellent account of 50% of the story, but you’ve left out 50%. That’s what I would argue. And, this is a central part of my theoretical work, and I have with Pascual Restrepo. And also with Pascual, we have also provided quite a lot of empirical evidence to bolster this empirical case, and other people have done as well.
So, you are absolutely right: If you introduce automation technologies, you reduce costs. Prices will decline. And, that may, under some conditions–if it’s sufficiently productive–also increase real wages. And it works exactly like you described, which is that: Eggs become cheaper, cars become cheaper, coats become cheaper, new widgets are introduced, and so on.
But, if that’s all that’s going on–automation and all that’s going on–labor’s share would decline steadily both at the industry level and at the economy level.
But, that’s not what we see in the data for much of the 20th century.
And the reason why labor’s share declines is exactly what you nailed, which is, you know, the employers–the machines–are the ones that are doing the egg production and the car production and the cloth production.
So, this is what Keynes also didn’t get. That’s why he was worried about technological unemployment.
And, what your story and Keynes’ story miss, is that at the same time–again, because we can apply scientific knowledge very broadly to very different things–we also create a lot of new tasks. Some of these new tasks are for capital. Some of them are for skilled labor. Some of them are centrally for unskilled labor.
And, what we document–and wonderful work by David Autor, Caroline Chin, Anna Salomons, and Bryan Seegmiller document even much better with much better data in a recent paper–is that throughout the post-war period, and in fact actually in the pre-war period as well, pre-World War II, concurrently with the mechanization of agriculture, there were a lot of new tasks for labor. Especially unskilled labor introduced in manufacturing and in other sectors of the economy.
And it is these new tasks that are so important that wages increase and employment is reinstated; and the labor share remains constant. And, that’s actually also very important for–
Russ Roberts: Where do those news tasks come from? They’re not springing up from–they’re not in the rain. They’re not exogenous. They’re endogenous—
Daron Acemoglu: Of course–
Russ Roberts: They’re part of the system.
Daron Acemoglu: They are 100% part of the system. And, again, some of the theoretical work that Pascual and I do was within this family of directed, technological change models where it is the entrepreneurs and the innovators motivated by profit that lead to these new tasks.
But, there is no necessity that the efficient balance of new tasks and automation will be created. Nor is that always going to be taking place at the same rate irrespective of the market structure.
So, perhaps when the tech sector–and now we can transition to AI [artificial intelligence] if you want–if the tech sector is monopolized by a few companies that may actually also affect the balance between new tasks and automation.
But, once we recognize that this balance between new tasks and automation is key–and that’s been the main thrust of my work with Pascual–that really puts a very different spotlight on how we should think about technological change. And we cannot just say, ‘Oh, technological change is ongoing. Everything is going to be fine.’
Yes, technological change is ongoing. That’s good. But, do we get the right balance between automation and new tasks? What are the market processes that are not working for that new balance to emerge? So, those become important questions to ask.
Russ Roberts: So, how would you possibly–let’s say you’re right. I don’t think it’s a balance, actually. I don’t think it’s like there’s a certain number of new tasks that pop up and a certain amount of unemployment or dis-employment that occurs from ongoing innovation. It’s a big, complicated soup, and they’re all tied to each other, and they’re all based on human creativity. But, let’s say you’re worried about it–as you are. How would you possibly think about how to get that balance right?
Would you stop certain automations from being–would you not allow–I mean, in one extreme, driverless cars if they ever came along, would be a really interesting example of this, right? Would throw millions of people out of work. And, I think that transition would be very painful, by the way. So, I think that’s an interesting tough policy question.
But, there’s no dial for balance between new jobs. So, what in practice would you suggest we could do to help people who are harmed by what I would call creative destruction?
Daron Acemoglu: Great question, Russ. And let me first preface by saying that in this, I am probably more aligned with you–and either the sort of Hayek perspective was already apparent in your question, is a decentralized system. Things are going to pop up because of their own decentralized dynamics. I completely agree with that. And I think it’s a fool’s errand for a bureaucrat or for the government to come and try to say, ‘You should innovate here and you should innovate there.’ I certainly do not advocate that. I don’t think it would work.
That being said, there are macroeconomic forces that could create bias.
Let me argue three of them are important. And I’m sure you’re going to disagree with at least some of them, but perhaps with all three of them.
One, is what the tax policy of the government does. So, if we make labor much more expensive and capital much cheaper for companies, then companies–being profit-maximizing entities–are going to be encouraged to use more capital, innovate more on capital, and not so much on labor.
And, what I do, for example, in a paper with Andrea Monera and Pascual Restrepo is compute what the tax rate on different types of capital and labor is–marginal tax rate–in the United States. And, the marginal tax rate on labor is about 25%, on capital is about 5%–equipment capital.
So, we’re creating a tax system. We have created a tax system that subsidizes people to do more automation and less of hiring workers. So, one thing to do is to make sure that we get rid of these distortions.
Second–I’m going from less controversial to more controversial. So, if you agree–
Russ Roberts: I’m not so sure, but go ahead–
Daron Acemoglu: Okay. If you agree that–and you haven’t, but some of your listeners may–that these quasi-rents are relevant, then automation also has a rent-shifting role.
So imagine that I was sharing some of my rents with my employees, and I was sort of happy because that was the equilibrium, but I also recognized that I’m creating a lot of welfare for them. The social planner likes that, because I’m giving them, like, a high wage. But, to me that’s a cost. And, now you give me an automation technology. That’s so-so. That’s what Pascual and I mean. It’s not that amazing, but it strips those rents away. I might be tempted to use it and the social planner doesn’t want me to.
So, that’s the second macroeconomic force.
And the third one is what we call vision; and that’s why vision plays such an important role in my book with Simon. The whole industry–the whole society–may get obsessed with a vision–temporarily, perhaps–that says, ‘Oh yes, we have to implement machine intelligence, artificial intelligence. Let’s get rid of humans.’ And, that can carry the economy for quite a while down an excessive automation path.
Russ Roberts: So, I just want to alert listeners: This conversation could go on for another hour, but because both Daron and I both face external constraints in the quasi-rents we’d like to enjoy, we’re going to have to cut it short in a little bit, which means we’re not going to get to ChatGPT and AI the way we might have. And, we’ll have to save that for another conversation, if you’re willing to come back, Daron?
Daron Acemoglu: I would love to.
Russ Roberts: Which is fantastic, because we’ve only done about 70 episodes in the last six weeks on AI, so it’s okay to have one that looks at these deeper philosophical issues. So, I’m okay with that. I hope we can come back to it though. You have a lot of interesting things to say. But I want to talk about these three and we’ll close with that.
So, on the [?coersion?]–on the taxation one, I would just observe that throughout, I would say most of all of the 20th century and a good chunk of the 21st, economists saw taxation of capital as bad for workers because they believed–you don’t–but they believed that in general capital made workers more productive and raised their wages. So, it was a good thing that there was that distortion.
So, it’s interesting that because of your view of the impact of technology on productivity, you’re taking a very different position. I just want to point that out.
But, the second thing that’s interesting is that I have often–I think on this program, even–argued that the minimum wage is an example that distorts the choices facing employers and pushes them toward replacing orders with capital. And I’m certainly against that; and I think that that’s a bad thing. We might debate over what level of minimum wage does that or how high it would have to be. But, in my view, especially for low skilled labor that struggles to find opportunities, I think minimum wage is a very mixed blessing. There’s, of course, a large empirical debate about that, that I would say not completely resolved.
What was the second factor?
Daron Acemoglu: Second is that if you are in an imperfect labor market with quasi-rents, then the social and the privately optimal impact of automation is different.
Russ Roberts: Yeah. And there’s no doubt that–I totally agree with you. That’s less controversial for me. The threat of technology–I would just add a couple things. One, as an employer–I’m not an employer–but as an employer who shares those rents–those gains–with workers, some of them might enjoy that. They might get pleasure from the fact that they provide employment at a nice wage and benefits for their employees and might find technology tempting. But, some would say, ‘You know, I just don’t want to put my workers out of work.’
Others, more, say, profit-oriented, money-oriented might be tempted. And, I certainly think that does affect the wages they offer. Actually I don’t disagree with that. So, I’m happy to say we’re on the same page there.
Daron Acemoglu: There, I think, there also has been other vision changes. The vision issue is not just confined to the tech industry. And I’m happy to talk about this in a different episode as well. The Friedman Doctrine–that the social responsibility of business is profits–in the abstract may appeal to economists, but I think it was part of a movement–and it was actually a motivator of a movement–for many managers to say, ‘Actually, let me think about this. Perhaps I shouldn’t really be so generous with my employees because that’s eating from the shareholders.’
I have some work, for example, with Alex He and Daniel le Maire showing how CEOs [Chief Executive Officers] with MBAs [Masters of Business Administration]–where ideas that shareholder value is key is quite emphasized–are very keen on not giving raises to their workers, whereas comparable companies do so. So, they have a negative effect on wage growth and labor share.
So, there are these other issues that intersect with exactly what you said, which is: What are the priorities of the CEOs? How short-term is[?ist?], versus long-term is[?ist?]? How much do they want to share the gains hoping that might perk at one point or that makes them feel better versus, ‘No, I have to go after my shareholders’ interests first and foremost’?
Russ Roberts: That’s very interesting. I would argue it may even be parallel–I’m very sympathetic to Friedman’s claim, but I take your point–and I think somewhat parallel to the economist’s relentless application of so-called rationality without thinking about a richer definition that somehow selfishness is always good because Adam Smith said so–which he didn’t–
Daron Acemoglu: Yes, he didn’t–
Russ Roberts: For starters. And, certainly there are many, many, many other things in our calculus besides money in terms of what we care about.
Russ Roberts: But, let’s close with some of the policy positions that you would think are important. I gave you a little bit of a hard time about there’s no lever to dial back certain things. But, presumably there are some general rules you might think about or have thought about for the concerns that you have. So, would you–the taxation would be one example: you’d want to make that different of labor versus capital. What else would you advocate in terms–
Daron Acemoglu: Let me talk about three others very briefly. And, in the book we try to clarify with, so let me even clarify. I don’t have the policy answers. What I hope the book achieves is that it fixes our aspirations on making technology more complementary to humans. And that I think is particularly important in the age of AI. And, if that happens, and people come up with other ideas for doing that–policy or otherwise–we are completely open to it.
But, let me give you, in that spirit, the other three policy applications.
The second is that we should have more sort of government competitions and programs to encourage human-complementary uses of AI and digital technology. The government, as I said, should not be at the driving seat, but the U.S. government through the Department of Defense, NIH [National Institutes of Health], NSF [National Science Foundation] in the past has had a very positive role in encouraging exploration of new areas. And, I think this is something that we should consider. Definitely not automation taxes, definitely not slow down automation. We do not want to slow down automation. We want automation to be rapid, but at the same time find other things that we can do for workers, so that worker productivity is also central.
Third, digital lab taxes. One area in which we think digital technologies, online interactions, and AI are being misused is through digital ads, which then creates this whole ecosystem of outrage, emotional engagement, extreme sort of reactions online. And, we think–there is a lot of psychological evidence; there is some political science evidence about how it changes this communication pattern–this is not a good use of how we use technologies through negative externalities, through emotional negative effects, especially on teenagers.
And that business model chokes other business models. Wikipedia doesn’t have misinformation. Netflix doesn’t have misinformation, but Facebook, Twitter, Reddit do. Why is that? It’s because of their business model.
So, we view digital attacks as a way of opening up new business models so that they can survive against free–‘we give you everything for free, but you are our data-model.
And then, the final one, most controversially, I think we should really go back to antitrust and rethink antitrust. We have never seen as large organizations in the world–private organizations–as Google, Microsoft, Facebook, Amazon. That means they have enormous amount of social, political, as well as economic power. Could that be justified because of the economies of scale and the competitive landscape? Or, should this be more of an antitrust in the same way that Standard Oil was targeted for antitrust? I don’t have the answer, but it’s really important that we have this conversation with it.
Russ Roberts: I would just encourage listeners–and you can close with the reaction of this, Daron–I just want to say for the record that in the other past episodes, I’ve mispronounced Daron’s first name. So, those of you who hear those and are wondering which one is right, it’s Daron [Da-ron’ with last syllable emphasized–Econlib Ed.].
I would just think about Schumpeter and I think about Capitalism, Socialism, and Democracy. I’m worried about some of the things you’re worried about with the current digital landscape. There’s many other things that worry both of us; I’m sure we just started scratching the surface.
But, I think it’s interesting to think about the fact that these large companies–you didn’t mention Apple, that would be another one–although they are large and although they have a lot of power within their fiefdoms–which would be, say, search right now in the case of, say, Google–all of a sudden ChatGPT is making them a little uneasy. It’s going to be an interesting situation.
So, I would just say, I agree that antitrust is woefully unprepared for this–is out of the picture, and maybe it shouldn’t[?] be. I’m not a big fan of antitrust, but right now, whether you like it or not, it’s not doing its–whatever you think it should be, it’s not doing it. But, it’s interesting to me that even in a time when it looks like these companies, inevitably we think, ‘Oh, they’re going to last forever. They’re so big. They could keep out competitors.’ All of a sudden Google looks a little vulnerable. Do you agree with that? Does that change your thinking on this at all?
Daron Acemoglu: I don’t. I don’t actually agree with that in the sense that I think right now the foundation models which–like ChatGPT-4–are so resource-intensive, and they’re going in a particular way that they might actually strengthen the hold of Google and Microsoft. And, OpenAI is no longer an open source nonprofit. It’s really part of Microsoft, and Microsoft is incorporating GPT-4 into all of its products. And, Google is really advanced relative to lots of other companies and foundation models as well. So, yes, I think AI broadly construed could be a competitive shock to that industry, but I think that’s not the path we’re going in right now.
Russ Roberts: My guest today has been Daron Acemoglu. His book with Simon Johnson is Power and Progress. Daron, thanks for being part of EconTalk.
Daron Acemoglu: Thank you for having me, Russ. This was fantastic.
